Why Microsoft's Special Unit Takeover of Inflection AI's Talent is Bad News for OpenAI
Also, GPT-5 is Coming.
Hey Everyone,
So when Microsoft developed a new unit called Microsoft AI, with two thirds of Inflection AI’s talent (likely the best part), it’s actually a bad sign for OpenAI.
The leaks of GPT-5 by coming to us by June, strikes me as an act of desperation by Sam Altman and company.
ChatGPT web growth has stalled. People don’t actually keep using ChatGPT, the wheels are falling off of the wagon of Generative AI hype, outside of Nvidia. Nvidia’s growth is ultra dependent on BigTech itself as the biggest customers. What could possibly go wrong?
GPT-5 Hype is all they got these days
OpenAI will have to release a lot of products, because most of what they have released isn’t sticking or providing much immersion for consumers.
It doesn’t matter how many speeches or podcasts Sam Altman goes. The GPT store is filling up with spam and traffic is starting to dry up. You can watch the entire interview with Lex here. Sam Altman always follows the same playbook.
Consumer AI demand isn’t really high, there's a now shift toward enterprise business models in AI. Given the real state of things, you can expect more gimmick product releases from OpenAI including Sora.
A report on Consumer AI by a16z found great uncertainty and transitory trends among consumers for Generative AI products. There was in short, a lot of turnover between their first list in 2023 and the one in 2024.
“over 40 percent of the companies on the list are new, compared to our initial September 2023 report.”
Consumer AI is a ‘Terrible Business’
While ChatGPT may be a winner, it’s not as big a winner as many are making it out to be. Microsoft is also noticing how Mistral, Anthropic and others are able to catch up to OpenAI’s tech, faster than some believed possible.
Microsoft had hoped to be years faster as a first-mover, that’s turning out to be months to the ever growing open-source Generative AI movement.
Last year, Garry Tan, one of the most successful tech investors in Silicon Valley, questioned the business case for consumer-facing AI technology like ChatGPT. He wasn’t alone, so did I and many people. Inflection being purged and raided by Microsoft means Pi may not have a bright future, and neither does Consumer AI in the nascent of Generative AI obviously.
No number of “leaks” by OpenAI or “speeches” by Sam Altman changes this reality. Microsoft understands its “productivity” suite of Copilots at least have a more business focus and while Github Copilot has obviously done incredibly well, that’s only after many years of development and Microsoft owning GitHub after all.
Will Microsoft’s other copilots do as well in the long-term? They are more Enterprise focused.
Consumer AI, from Bing AI to Google Gemini, i.e. “Bard”, have been a disaster and consumers simply don’t trust them.
OpenAI is expected to release GPT-5, an improved version of the AI language model that powers ChatGPT, sometime in mid-2024—and likely during the summer. So GPT-5 can trend on Twitter again, but not all is right with the future. OpenAI recently did a tender off valuing the company at over $85 Billion dollars. While revenue growth has been doing good, how about expenses, all that talent and all that compute required?
How about the efficacy of OpenAI to meet the actual needs of the $13 Billion investment from Microsoft. Now it cannot be that good if Microsoft needs to break apart Inflection just to make its own “Microsoft AI” unit.
Microsoft is building a New Consumer AI Unit Because OpenAI is Doing Poorly
To pillage Inflection means Pi had no future. It means OpenAI isn’t doing it for Microsoft any longer, just a short time after its massive investment in the much-hyped startup.
This week, startup Inflection AI partially imploded, or whatever you want to call it. The startup's two main founders decamped to Microsoft, taking a sizable team with them. Microsoft AI, is likely just getting started. Instead of the funding whiz kid Sam Altman, they got Mustafa Suleyman, someone who was marketing his book more than his startup last time I checked.
Consumer AI really isn’t a great story. Cohere that wants even more funding, makes all of $12 million a year. It’s insane. They too, want a $5 Billion valuation, eh? The OpenAI of Canada doesn’t look so great.
A bunch of Billionaires huddled together to fund Inflection, the biggest head-fake like HER we’ve yet to see, and what a deception. Inflection raised more than $1 billion last year at a $4 billion valuation, with backers including Bill Gates, Eric Schmidt, and Nvidia. Most of that destroyed the value of the company and went to compute and those H100s (Nvidia thanked Pi very much).
They did this all the while Reid Hoffman was on the Board of directors of Microsoft. It doesn’t seem like he was a “silent observer” on that particular day. But what people aren’t talking about is it means OpenAI aren’t any longer the exclusive favorite child of Microsoft. There’s Mistral, there’s others for Azure, there’s not a huge trust even by the business world in Microsoft 365 Copilot. There is a feeling it was all rushed.
GPT-5 is in the works for sure, some enterprise customers have recently received demos of GPT-5 and related enhancements to ChatGPT, but does it even matter? This isn’t the same world as when ChatGPT first came out or even GPT-4 came out. It’s already a crowded market with Anthropic, Mistral and others closing in. I wouldn’t even be surprised if China closed in, as soon as in 2025.
OpenAI never had a monopoly on any of this and Microsoft’ first-mover advantage is going to fade pretty badly. Whether that situates Bing better or an OpenAI search of its own (based on Bing), remains to be seen. Certainly Azure has grown faster with the hype, than competitors AWS or Google Cloud. But certainly nothing as good as Microsoft was claiming would happen.
Some Basic Facts about Generative AI in 2024
Consumer AI is not a mature market
Generative AI first wave products are lacklustre
Most people mistrust AI products made by BigTech or well funded startups
Microsoft moved too fast and the first impression is not great even for Enterprise clients.
OpenAI has not been a stellar partner for Microsoft
Inflection's main product was Pi, an AI chatbot for consumers that provided emotional support and advice. Pi felt like Claude but it used a lot more emojis. Pi was charming, if AI companions were our thing, and if you were a woman in your 50s in North America or a European country. Certainly some form of Pi will continue to exist.
But Generative AI is woefully immature. It doesn’t matter how many companies use AGI in their marketing. And a lot of the AI prophets of this era won’t even lead startups that are profitable, ever.
It didn’t matter how many Twitter followers Sam Altman got, or what he said. You sort of knew how this would all play out in the end. The other version of OpenAI and model makers of the early days, likely won’t make it either.
Consumer AI is a terrible way to make money.
Enterprise AI is a very difficult way to make money and win big clients and keep customers. Pure-play AI companies rarely do anything business wise. Open source AI can already replicate most of what these companies can offer in 2024.
GPT-5 Won’t Save the Microsoft Partnership for OpenAI
OpenAI have more of a chance of getting funding from Saudi Arabia via a16z than getting a lot more funding from Microsoft.
The Middle Eastern country is creating a gigantic fund to invest in A.I. technology, potentially becoming the largest player in the hot market, to the tune of $40 Billion, which isn’t Softbank Vision fund scale, but it’s decent.
While the U.S. gives money away to Intel, and a horrible idea in Phoenix, Generative AI cannot possibly measure up to the expectations it created for itself. A lot of this comes down to the “AGI” marketing chosen by Sam Altman and his rich friends. But that’s also just the path that destroys trust the most with your average real-world skeptical and nihilistic consumer.
Sam Altman said OpenAI is not ready to unveil what Q* is or details about this mysterious initiative. Between Sora, GPT-5 and Sam Altman’s behavior, this is getting very tiresome.
So what’s the truth in all of this? ChatGPT is stagnating. The megahit chatbot from OpenAI has seen declining web traffic in five of the past eight months and is currently down 11% from its May 2023 peak.
The hoax that ChatGPT is somehow useful or a must-have is losing momentum. It’s not even a unique interface, chatbot or advantage for most jobs or most people. It has limited entertainment value, no matter what use case you can spin out of it. The reality is ChatGPT and all of these chatbots and copilots likely don’t have a great future if tech history is any indication.
Generative AI is developing rapidly enough that OpenAI might profit more from a Search product or from Sora, more so than ChatGPT itself. OpenAI may not even make it, as new companies and new products come to the forefront.
Sam Altman is not the messiah, he’s just a dude that likes to invest in a bunch of things and pretend his startup are the leaders who will bring the world AGI. Whatever that even means.
Read more:
https://www.nytimes.com/2024/03/19/business/saudi-arabia-investment-artificial-intelligence.html
https://a16z.com/100-gen-ai-apps/ (the March, 2024 version)